ueaj

Researcher @pangramlabs -

[ueaj @_ueaj]

Normally I wouldn't say anything at this stage but I feel that the moment calls for it. This isn't directed at any particular company (I'm serious), but it's something we should all start thinking about now.

I'm moving to NYC to join @pangram

The midterms are coming up, and I'm worried about AI generated misinformation and other misuses of AI. Even right now the bots are on overdrive bc of the war, this is bad.

My old job paid ~210k/yr, more money than I knew what to do with. (that's also why no new blogs/research)

I don't expect to stay at pangram long enough to vest, and so in all likelihood I will be taking a 40k/yr pay cut to do this. And of that income I intend to donate a good amount (political) too. I will be leaving all my friends here. I've gotten vastly better at making new ones but is still very uncomfortable. And I will be going in person instead of working online. (None of this is pangram's fault btw I intentionally didn't negotiate, and they're aware of all the details)

My questions to you are:
1. How much is your conscience worth?
2. How easily can your conscience bend?
3. Will ambiguity, intentional or not, satisfy you?
4. Will you simultaneously be capable of good compromise when the time demands it?
5. Can you, right now, in good faith, say that what you are doing is right?

Our society is falling apart at the seams, trust is degrading, loneliness is at an all time high, antisocial behavior is common, gambling is everywhere, influencers sell out the truth and their viewers for money.

We are building an extraordinarily powerful technology while it all happens.

Are you prepared to be the people the world needs right now?

[ueaj @_ueaj]

I don't think crypto is doing very well frankly, nor did it really replace any firm, nor would it, as the reasons for financial regulation are learned lessons (i.e. the non-market intelligent system of the government).

The markets are bigger because the markets generally encompass the ever larger non-market systems. But the systems that enclose markets are also getting more powerful too, like governments. Imagine a system where there are 2 companies, if the 2 companies control everything and operate internally as dictatorships, did the market get bigger or centralized control?

Obviously centralized control, and again we can calculate this exactly / unambigously by observing the change of power laws over time, they almost monotonically increase in inequality not decrease. This alone should be a sufficient argument to disprove it imo

I wrote an essay getting into this a little bit, but my challenge to you is, if market systems are actually *generally* most efficient/intelligent, then where is AGI? If it were that simple then go glue together a big multi agent market system and make a trillion dollars.

[ueaj @_ueaj]

I think this starts from the humanist POV, which is that the individual agents are unpredictable or have free will. Bargaining costs can't go to zero, but the floor on the costs is not set by individual actors but by mutual information.

The way to reduce costs is computational reduction, which is most optimal when it is done across multiple agents (i.e. coordination).

This is why for example it's more efficient to finetune an LLM trained on everyone else's data than it is to train it solely on yours. It's not generally possible to reduce bargaining costs below that which can be done with coordination.

[ueaj @_ueaj]

"larger social systems allows markets where they were previously too costly to organize" in principle the incentives always existed, there just wasn't institutions intelligent enough to respond to them.

It's not that new markets are being *created* in response to the formation of these higher order institutions, it's that there is some finite number of them that are being planned away one at a time.

Eventually there will no longer be major outside threats to induce market selection, unless aliens are real and we have to compete against other ASIs. (i.e. if there is an infinite nesting of brutal fast acting incentives) This might be true but I don't think they'd be fast acting, i.e. humanity could enjoy several trillions/billions years of peace before having to engage in a race to the bottom w/ thermodynamics or aliens

[ueaj @_ueaj]

@dnqwnscASd @BIMBOSATTVA_ @1thousandfaces_ I don't think this is true? Transaction costs have been going down with coordination and intelligence and this has produced more centralization and tighter power laws not less

[ueaj @_ueaj]

@dnqwnscASd @BIMBOSATTVA_ @1thousandfaces_ The scope of the firm is ever increasing though, states don't operate as markets, and more powerful non market systems are being made. And importantly they grow to *replace* what was once a market based system.

[ueaj @_ueaj]

Yeah but we can quantify the balance between these two forces (centralization efficiency vs higher order markets) by measuring how power law have changed over time, and it almost monotonically moves in favor of centralization.

The higher order markets always existed, independent of whether the "tier below" had been socialized, it's not that there are more higher order markets being created as we become more advanced, but a finite set of them that are being planned away one at a time.

[ueaj @_ueaj]

This would make sense if the world was becoming more decentralized, but it's not. Monopolies form naturally, inequality increases, empires grow, it's very obviously the case that the market is generally not optimal given that in nearly every area of endeavor it's replaced as soon as the technology is available.

[ueaj @_ueaj]

@dnqwnscASd @BIMBOSATTVA_ @1thousandfaces_ Sorry, correction, the system you are describing is one *possible* global system, but it's not likely to be the best or inevitable one and regardless it is not possible without governments and so on. It would likely look more like an empire or world government

[ueaj @_ueaj]

> Essentially, it's not true that a market is the always most efficient organizational structure, it is only effecient when there is not sufficient intelligence and coordination, which has to be built slowly and over time.

This is the TL;DR and it's the coase theory of the firm. The system you are describing is perhaps inevitable, but requires a level of intelligence and maturity that will take time to achieve, it is not possible right now.

[ueaj @_ueaj]

As the world has become more intelligent and interconnected we have become more inequal and more monopolar, not less. The roman empire, firms, all prior empires, etc. were all limited in size and time by interconnectivity, navigation technology, etc., and as it increased the empires became bigger.

A market system is only efficient when you don't have the intelligence, communication and coordination to produce a more efficient system. But as we become more intelligent, the less need there is for multipolarity and competition.

[ueaj @_ueaj]

If free will and rationality isn't real -> people aren't rational -> people are not generally capable of acting in their higher order interest -> you need republics and firms to perform selection on high order intelligent systems to respond to high order incentives -> you need governments

[ueaj @_ueaj]

Constructing such a system is not in the low order interest of any individual, and because individuals can't directly observe or act on high order incentives we can't construct such a system. It is evolutionarily selected for, in that it is more economically effecient, and thus prototypal systems like globalization and free trade will outcompete those that don't. Eventually we will arrive at something like what you mention but that system is monopolar, not multipolar, and involves creating systems, etc. that are able to optimize for high order incentives.

i.e... governments again

[ueaj @_ueaj]

@emmetics I'm tired boss

[ueaj @_ueaj]

"Attention is just a special case of <abstract math thing> so we generalized it by <neglecting the other 30 abstractions and conditions required for frontier architecture> and we found it performed <p hacking> compared to <naive baseline>"

[ueaj @_ueaj]

@dnqwnscASd @BIMBOSATTVA_ @1thousandfaces_ this is just defacto multipolarity (i.e. what we are close to right now) and

1. it doesn't work and isn't a stable equilibrium
2. leads to immense suffering, wars, starvation and is bad

The multi country system with the occasional war *is* market competition on governance

[ueaj @_ueaj]

@zeta_globin this is just twitter

[ueaj @_ueaj]

@BIMBOSATTVA_ @1thousandfaces_ there's probably some overlap but no I think the libertarianism is just convergent evolution

[ueaj @_ueaj]

@credenzaclear2 this is just the giving pledge

[ueaj @_ueaj]

Not strictly but libertarianism in general is a cognitohazard for autistic people, as they are disproportionately rejected by society and thus resent authority or prosocial behavior in general.

Rationalists are just disproportionately autistic, but not capable of overcoming their biases to recognize people are not actually independent creatures but require care and protection.

They're generally intelligent enough to know free will isn't real but if you dare say the average joe shouldn't have access to sports betting or peptides this is like a crime against freedom or something.

[ueaj @_ueaj]

wow you're telling me your solution that happens to solve attention sinks outperforms a baseline without an attention sinks patch? damn I guess the optimal transballs theorem is really AGI then better write a 10,000 line kernel for this specific architecture